Segmentation of unconstrained handwritten sentence images using a time delay neural network

초록

An approach for locating word breaks from a line of totally unconstrained handwritten text image is presented in this paper. Clues, which reflect the authors writing style in terms of spacing, are extracted from the, major vertical strokes and interpreted using a time delay neural network. Role of vertical strokes in a script is emphasized. As a result, the approach ran be regarded as a unified means for combining the character segmentation and the word segmentation. To prove the effectiveness of the proposed approach, comparisons with. other conventional connected component based approaches were made.

키워드